Traditional Governance, Citizen Engagementand Local Public Goods: Evidence from Mexico!

Alberto Dıaz-CayerosBeatriz Magaloni

Alexander Ruiz Euler

Abstract

In this paper we explore the role of traditional governance in theprovision of public goods in indigenous communities in Mexico. Ourtheoretical approach proposes that poor communities that are governedthrough direct participatory democracy practices based on indigenoustraditions will be better able to provide local public goods than commu-nities that have delegated to political parties collective-decision making.We propose three specific channels through which traditional governancea!ects local public good provision: community headmen’s institutionalembeddedness, broader civic engagement in collective-decision making,and social sanctions. We empirically assess our theory with a two-pronged strategy to analyze observational data, controlling for selectione!ects. Our results reveal that in traditional governance municipalities,citizens are significantly more engaged in collective decision making,and as a result, the intra-community distribution of social services, in-cluding potable water and sewage, exhibits less anti-poor biases thanin municipalities ruled by political parties, which are subject to elitecapture.

1 Introduction

The goal of this paper is to assess the e!ects of traditional governance on local

public good provision. We ask if poor indigenous communities are better or

!This research has been made possible by the generous support of Stanford University,the University of California O"ce of the President (UCOP), and the Center for US-MexicanStudies in the University of California, San Diego (UCSD). A very preliminary version ofthe research project was presented at the Initiative for Policy Dialogue Decentralization TFMeeting on 10-11 June 2009 at Columbia University and at the Faculty Research Seminarof the School of International Relations and Pacific Study (IRPS) at UCSD. We thankcomments and suggestions by all participants, and in particular Vidal Romero and Jean-Paul Faguet. Of course, all errors remain our responsibility.

1

worst o! by choosing to govern themselves through customary law and par-

ticipatory democracy, versus delegating decisions concerning the provision of

public goods to political parties. The literature o!ers primarily three expla-

nations for the di!erential provision of public goods. First, ethnic diversity

seems to make public good provision more di"cult. Second, governments that

in some measurable way are more capable might be better able to provide

public goods. Third, local government accountability and civic engagement,

particularly by women, are essential to curb rent seeking and capture in the

provision of public goods. Our paper seeks to contribute to this body of re-

search by focusing on the e!ects of local governance on public good provision.

Our research design takes advantage of an important institutional innova-

tion in the state of Oaxaca, Mexico, that in 1995 allowed indigenous commu-

nities to choose between two forms of governance: on the one hand, a form of

“traditional indigenous governance” called usos y costumbres (usos hereafter),

which entails electing individuals to leadership positions through customary

law in non-party-sponsored elections, making decisions through direct partici-

patory democracy, and monitoring compliance through a parallel system of law

enforcement and community justice; on the other hand, “party governance”,

which entails the selection of municipal authorities through electoral competi-

tion among political parties and adjudication of conflicts through existing insti-

tutional channels, including the state and federal judiciary. This legal change

provides a unique opportunity to make inferences about the e!ects of local

participatory democracy vis-a-vis political parties. Although municipalities

di!er across type of governance, they share formal political institutions such

as the structure of municipal government (e.g. having a municipal president)

or the legal and fiscal relationship to the state and the federation. We have

thus variation in governance type while retaining the structure of government

2

fixed. Our theoretical approach proposes that poor communities that are gov-

erned through direct participatory democracy practices based on indigenous

traditions will be better able to provide local public goods than communities

that have delegated to political parties collective-decision making.

We propose three specific channels through which traditional governance

a!ects local public good provision: community headmen’s institutional em-

beddenes, broader civic engagement in collective-decision making, and social

sanctions. We argue that traditional governance practices, including direct

participatory democracy practices, citizens obligation to provide services for

the community, and the establishment of a parallel system of justice, better

allow poor communities to prevent elite capture and to monitor and sanction

free-riding.

Before 1995 municipalities in Oaxaca were formally governed by political

parties. The choice that municipalities faced after the constitutional reform

was whether they wanted to maintain a party-based system of governance, or

switch to usos. Our research design involves a multi-method approach first

using aggregate data measuring household coverage of local public goods and

municipal poverty, on the one hand, and individual-level data measuring civic

engagement in collective decision-making and intra-community distribution of

public services, on the other.

Aware of the fact that there may be strong selection e!ects, we use a quasi-

experimental design to estimate the e!ects of usos y costumbres on aggregate

indicators of public service coverage. We first instrument for the self-selection

of municipalities into governance type, and use the estimation of the probabil-

ity of selection as a propensity score to create two sets of comparable (cuasi)

treatment and control groups: one for the 1990 census, and one for the 2000

census. In order to control for fixed unobservables and estimate the average

3

treatment e!ects on the treated (ATT) we use a first di!erences approach, mea-

suring public good provision as the change in levels before and after the 1995

reforms. Our results show that electricity provision increased faster in those

municipalities governed by usos. They also suggest that traditional gover-

nance municipalities experienced more significant reductions in poverty. With

respect to citizen engagement and elite capture, contrary to existing scholarly,

we find no evidence of entrenchment of local bosses (caciques) associated with

the former ruling party, the Institutional Revolutionary party (PRI), in places

ruled by usos. Moreover, on average municipalities governed by political par-

ties are less likely to hold open council meetings wherein citizens participate

in decision-making processes.

We attribute better electricity coverage and faster poverty reduction to dif-

ferences in governance and collective decision making practices shaping intra-

community distributive politics, wherein there is significantly less bias toward

the poor and uneducated in usos municipalities relative to party ones. To sup-

port our conclusions derived from aggregate-levle data, the second stage of our

empirical approach involved the design and collection of a stratified random

sample involving 600 questionnaires of men and women over 18 years old in

rural and semi-rural areas in Oaxaca. The stratification divided Oaxaca mu-

nicipalities according to the size of their cabecera (the core town, seat of the

municipal government) and their governance institutions. The survey asked

respondents about their participation in collective decision making and their

access to public services. Although water, electricity, and sewerages are in

theory public services, access to these services is not universal. For example,

electricity might exist in a village, but this does not mean that all streets will

have public lighting, or that all households will be connected to the public grid.

Households in the outskirts of the villages might not be reached. Similarly,

4

public goods vary in their quality although the municipality might invest in

road pavement, some streets might have more wholes than others. Water is

scarce in the villages that we visited: even when a household in theory had

access to potable water, often this arrived only once a day or even once a

week. Our survey allows us to model intra-cimmunity distribution of these

public services by asking respondents about their access to them.

To model civic engagement and intra-community service distribution, we

implement a Heckman correction model to adress the problem of selection into

either type of governance astructure (Heckman, 1979). A Heckman correction

allows us to take into account the potential selection bias that may alter the

estimates, and by doing a separate regression for each governance type using

the appropriate inverse Mills ratios, to allow variation in the regression coe"-

cients by governance type. This enables us to explore the e!ects of governance

type on each of these coe"cients.

Our main findings show that even after controlling for selection e!ects,

there is considerably more civic engagement in usos municipalities while cit-

izens remain profoundly apathetic, uniformed, and disengaged in municipali-

ties governed by political parties. Our results further indicate that collective

decision-making in party municipalities tends to be captured by a small party

elite, which distributes public services –water and sewage– predominantly to

wealthier individuals and those who belong to the party’s network or political

clientele. By contrast, participatory democracy practices in usos municipali-

ties permit a broader group of citizens to engage in collective-decision making

and as a result we find a more egalitarian distribution of public services.

This paper is structured as follows. The next section introduces the scope

and limitations of existing hypotheses on the e!ects of local governance on the

provision of public goods. The third section presents a typology of traditional

5

and party-based governance systems in Mexico. We then describe our data

and elaborate on the methodology used to create the counterfactuals to assess

the e!ects of governance on public good coverage using aggregate data. We

proceed then to describe our survey and present our empirical results regarding

civic engagement and intra-community public service distribution. The last

sections discuss our results and conclude.

2 Traditional Governance

Understanding the e!ects of traditional governance on public good provision is

important both from a theoretical and policy perspectives. As noted by Besley

(2006), a general shortcoming in development research is that we know far less

about local public good provision than about policy interventions aimed at

income support, even though the former have equally important e!ects on well-

being. Well-being hinges not just on individual income, but on access to public

goods and services such as potable water, sewerage, electricity, schools, and

health clinics. Furthermore, a better understanding of governance and public

good provision in highly marginalized villages brings us closer to addressing

crucial issues about poverty and improvements in material well-being.

The provision of public goods and its relationship to governance has re-

ceived increased attention by the academic and policy communities. The most

influential hypothesis in the last few years has been one associated with Alesina

et al. (1999), which proposes that greater social heterogeneity –as measured

through an index of ethnolinguistic fractionalization (ELF)– makes it harder

for communities to provide public goods. Such failure is attributed to the

idea that it is more di"cult for groups to engage in collective action when the

marginal utility of a public good di!ers across sub-groups. In Alesina et al.

6

preference heterogeneity is driven by ethnic fractionalization. Studies finding

evidence of the impact of social heterogeneity in public good provision across

nations and within countries include Alesina and La Ferrara (2000); Khwaja

(2009); Miguel (2004); Miguel and Gugerty (2005); Dayton-Johnson (2000);

and Baqir (2002).

The most recent research agenda on the role of ethnic diversity and public

good provision moves away from cross-sectional variation to a focus in local ex-

perimental settings, in which scholars have tried to understand the conditions

under which communities are more able to create networks of trust. Hab-

yarimana, Humphreys, Posner, and Weinstein (2007) in particular, performed

experiments in Kampala, Uganda, testing the willingness of co-ethnics and non

co-ethnics to cooperate. Their results show that co-ethnics cooperate more,

and they attribute this finding to the existence of denser ethnic-based insti-

tutions that allow for monitoring and sanctioning of non-cooperative behav-

ior. Confirming older psychological findings, recent economic experiments also

show a higher level of pro-social behavior within ethnic groups than between

groups (Bernhard, Fischbacher, and Fehr, 2006; Glaeser, Laibson, Scheinkman,

and Soutter, 2000).

In this literature social structures are the main explanation of failures in

public service provision, disregarding other potential causal links such as elite

capture and government accountability. There is a long tradition of research,

especially among political scientists, which has seen public good provision

through the lens of state capacity (Kohli, 2001). The general thrust of that

literature has been to suggest that failures in the provision of public goods re-

flect underlying problems arising from weak states that are incapable of taxing,

running a bureaucracy, or in general, fulfilling basic public functions. How-

ever, when it comes to unpacking state capacity –its elements, causes, and

7

consequences– the literature remains quite underdeveloped1.

Policy makers have increasingly paid more attention to local power struc-

tures and corruption as explanations for the di"culties governments face in

providing public goods and services2. In a particularly poignant example,

Reinikka and Svensson (2004) measured an astounding leakage of 87 percent

in a program in Uganda meant to provide grants to schools for non-wage

expenditures. Such leakage was successfully reduced through greater citizen

involvement and information regarding the allocation of funds to the local

schools (World Bank, 2003)3. Olken (2006) similarly found that the leakages

in a poverty relief program delivering rice in Indonesia were large enough that

they o!set the welfare gains from having the program in place at all.

As a result of these findings, many scholars are now pointing to both top-

down and bottom-up accountability mechanisms and their impact on local

public good provision such as auditing to local o"cials (Ferraz and Finan,

2005); providing citizens with increased access to information about govern-

ment performance; and community monitoring at the local level, especially of

service providers.

The dynamics of party competition also seem to matter. Chhibber and

1For example, Acemoglu, Johnson, and Robinson (2001) deployed an instrument for goodinstitutions (settler mortality rates) rather than any direct measure of it. In Fearon andLaitin (2003) the conditions that favor insurgency are linked to the strength of the statebut the authors use a set of proxies (e.g., GDP/capita, mountainous terrain, reliance on oilrevenues) instead of measuring state strength directly.

2Bardhan and Mookherjee (2000) have shown, in a formal model, that centralized systemsof public service delivery are more subject to corruption. However, they also note that localelites might capture governments making them less e"cient than a centralized arrangement.Besley and Coate (2003) have provided a model in which the advantages of decentralizationdepend on legislative behavior and how jurisdictional spillovers and conflicts arising fromthe variance in preferences over public good provision across places are mediated by thepolitical system. Despite these theoretical advances, we are only starting to understand thelinks between democratic accountability, local public good provision and decentralization.

3In an empirical evaluation of Sen’s (1981) influential hypothesis that democracy preventsfamines, Besley and Burgess (2002) have shown that Indian states with greater freedom ofthe press are more likely to deliver disaster relief. Besley and Prat (2001) have similarfindings for a cross section of countries.

8

Nooruddin (2004) found that Indian states where patterns of electoral com-

petition are sti!er, in the sense that the incumbent faces strong contestation

from one single challenger, are more likely to provide public goods. Diaz-

Cayeros, Magaloni, and Estevez (2007) also demonstrate that there is more

investment in public good provision relative to clientelistic or particularistic

transfers in more competitive races in Mexico. Cleary (2004) does not find

this e!ect of electoral configurations on public service provision, although he

shows that variables related to political participation, such as literacy and

turnout, improve public service delivery4. Closely related to the argument of

participation, in the case of India, Chattopadhyay and Duflo (2004) have ar-

gued that female leadership in Village Councils led to greater investments in

infrastructure related public goods (water, fuel and roads), while men tend to

favor investments in education.

Our paper seeks to contribute to this body of research by focusing on the

e!ects of local governance on public good provision. Our empirical analysis

builds on ethnographic and aggregate data of the southern state of Oaxaca in

Mexico (one of the most impoverished and indigenous in the country), which

constitution allows municipalities since 1995 to select among two very di!erent

types of governance structures –those we have referred to as traditional indige-

nous and party governance. We ask whether communitarian self-rule improves

local government accountability and the provision of local public goods, or

as the conventional view holds, represents a form of autocratic enclave that

works to protect entrenched local power structures and caciques allied with

the former ruling party.

For most of Mexico’s modern history, forms of indigenous communitarian

self-rule have been extra-constitutional, although it has been widely practiced

4Platteau (2004) criticizes, however, participatory models of local level development.

9

and tolerated de facto by the federal regime (Cleary, 2007). In the mid-1990s,

the state government of Oaxaca formally recognized usos y costumbres depart-

ing from constitutional doctrine and practice dating back to the 19th century,

through which state authorities forbade indigenous communities from exercis-

ing formal autonomy over local political a!airs along with the ability to use

their own chosen institutions of governance.

Existing evaluations of this form of governance by political scientists are

unfavorable. The prevailing view is that the reform in Oaxaca attempted to

limit the extent of the anti-PRI vote (Benton, 2008) and insulate local PRI-

supporting indigenous caciques by taking partisan elections o! of the table

entirely (Cleary, 2004). The argument is that the PRI allowed indigenous

communities to choose usos in order to prevent opposition political parties

from entering these impoverished political markets. In impeding opposition

parties from competing in these elections, so the argument goes, the PRI was

better able to sustain its political monopoly in the state of Oaxaca. Scholars

also have found that this form of indigenous self-rule is highly discriminatory

of women (Eisenstadt, 2007).

A key limitation of much of the existing research is that it ignores the prob-

lem of selection bias. Since the presence of usos y costumbres is non-random5,

it is not possible to claim any causal e!ects without properly addressing the

problem of self-selection first. For example, when scholars argue that tradi-

tional governance works to protect the monopoly of the PRI, it is not clear

if these municipalities exhibit significantly less political competition (as mea-

sured by vote shares in federal elections as in the existing scholarly work)

because the PRI purposely designed usos to take the opposition out of the lo-

5The Oaxacan congress passed a law allowing for election by usos y costumbres in Septem-ber of 1995, and in the municipal elections held two months later, 412 municipalities choseto select their leaders under this laws provision (and as the details of the law were ironedout, six additional municipalities joined).

10

cal political landscape or because these communities are intrinsically less likely

to divide along partisan lines. Since random assignment of treatment is impos-

sible in our setting, we work with observational data and use a multi-method

approach to address the problem of selection bias.

3 Variation in Local Governance and its Impact on Lo-

cal Public Good Provision

All municipalities in Oaxaca –regardless of whether they are ruled by parties or

usos– share the same formal institutional structure: they all have a municipal

president and a set of aides whose attributions stem from the federal and local

constitutions, and they all share a similar legal and fiscal relationship with their

state and the federation. Importantly, there is a constitutional prohibition for

reelection in Mexico, which means that municipal presidents cannot run again

for that position in the consecutive term.

Municipal presidents are in charge of making decisions over investments

in local public good provision or social infrastructure projects that the con-

stitution reserves to this level of government (roads, markets, public lighting,

sewerage, street pavement, granary, slaughter houses, and the like). In poor

municipalities, funds for investment in social infrastructure overwhelmingly

come from federal transfers within the FISM (Fondo de Infraestructura Social

Municipal), which is governed by a formula that is based on poverty indica-

tors. These transfers are given by the federation directly to the municipalities,

which retain ample discretion with regards to how these funds are distributed

and used, the only restriction being that they need to be spent in social in-

frastructure projects. Federal transfers for social infrastructure projects within

the FISM have significantly increased since 1997, allowing municipal presidents

11

to play an increasingly more important role in the provision and distribution

of local public goods. Thus, for the purpose of this analysis we can take as

exogenous the resources municipalities have at their disposal.

Despite an identical formal institutional and fiscal setting, there are impor-

tant informal institutional di!erences among party and usos municipalities.

We highlight the following key di!erences:

3.1 Selection of political leaders

Party municipalities select leaders through partisan elections. Political parties

channel the progressive ambition of these politicians; solve collective action

dilemmas and bargaining among di!erent levels and branches of government;

and hold their politicians accountable through the threat of loss of future utility

streams through the assignment of o"ce. The rule of no-reelection implies

that municipal presidents interested in a career in politics need to service the

leadership of their parties rather than their constituencies to get promoted

to higher o"ce. Governors are often the key figures determining who gets

promoted. In the state of Oaxaca, the long-lasting PRI had been the dominant

political party and had never lost the governorship of the state –until the 2010

gubernatorial election. The absence of consecutive reelection has generated a

very deficient accountability mechanism at the municipal level in Mexico. The

perverse system implies that if their parties decide to renominate irresponsive,

incompetent, or corrupt o"cials for higher o"ces, there is virtually no way

for the electorate to directly punish municipal presidents who underperform

or even steal government funds.

Usos municipalities choose representatives in assemblies. Citizens meet

in a public assembly, deliberate, and vote. There are no uniform rules with

respect to how this collective decision-making should take place. In some

12

municipalities, the main neighborhoods nominate a number of candidates for

the o"ce of the mayor, and the full assembly then discusses each candidate

until they are able to whittle the list down to four or five. “They then vote

’por terna’, which is a form of approval voting in which the assembly considers

one candidate at a time and counts a"rmative votes. The candidate with

the most approval votes wins, and serves a single three-year term” (Cleary,

2008: 15). In other municipalities, the serving municipal president proposes

the names of three or four candidates, and the assembly then decides. In yet

others, potential candidates must come from those who have a long and well-

established record of serving the community. For the most part, women are

not considered.

Voting can be secret or not. Most of the time a long deliberation takes place

that can last up to ten hours and citizens raise their hands to vote. Formally, all

such elections are supposed to be non-partisan, and the parties are not allowed

to support candidates or run campaigns; in practice, however, political parties

are sometimes associated with some candidates. A key di!erence in the usos

municipalities is that municipal presidents are members of the community, and

although these can’t get reelected either, most of the time they will continue to

live and to participate in collective decision making even after they step down

from their posts. Moreover, our ethnographic research reveals that those who

get elected to the municipal presidency normally have a record of servicing

the communities in previous “cargos” or services. Municipal presidents in

Usos municipalities are hence generally better embedded in their communities,

which we believe, contributes to generate better accountability mechanisms

than in party municipalities.

13

3.2 Solution of Collective Action Dilemmas

People in municipalities appoint local authorities in elections every three years.

Once in o"ce, executives usually make political decisions without much public

scrutiny. Formally,the municipality is composed of a president (the executive)

and a separate ‘‘cabildo” (assembly), although by design the largest party has

the majority and the government is always unified. The usual monitoring and

sanctioning mechanisms are performed by political parties.

In usos municipalities, collective decisions are taken item by item, mostly

in assemblies where there is a process of collective deliberation. Assemblies

meet regularly and citizens are allowed to participate; they are usually well

attended and can often take hours. This publicness of political assemblies

reinforces the collective monitoring mechanisms that elsewhere fall on political

parties. Traditionally, women were excluded from participating in assemblies6.

3.3 Taxation

In party municipalities, taxation is mostly exogenous to the provision of public

goods: political leaders decide tax bases and rates, and investment decisions

are often made independent of these. By contrast, in usos municipalities, the

community decides tax structures, which include non-pecuniary contributions

that are endogenous to the level of public goods provision. For example, when

the community decides that a road needs to be repaved, or that garbage needs

to be picked, some or most members of the community will be asked to pay with

their own labor –a practice called “tequio”, which is usually not remunerated.

6Although our field research revealed that this practice has changed in the last years,with women being allowed to participate and vote in most places.

14

3.4 Servicio

A key di!erence between party and usos municipalities relates to public jobs

and “cargos”. In party municipalities, all public services are performed by

municipal employees, who are remunerated with public funds. Political parties

have ample leeway to distribute these public jobs to their clienteles. In usos

municipalities, members of the community are expected to perform public

services, and with the exception of the municipal presidency, these services

must be provided, often without remuneration. Citizens are hence deeply

involved in the life of the community, either organizing religious festivities,

coordinating public works, cleaning the roads, policing the streets, rural roads

and fields, repairing the school, organizing the public market, decorating the

church, etc7.

3.5 Monitoring and Sanctioning

Party municipalities have a professional armed police force. Conflict resolution

and adjudication takes the regular judicial channels, which are often located

outside of the communities, and are administered by judges that have little

knowledge of the locality. Usos municipalities have a parallel system of conflict

resolution and adjudication and a parallel security system. The community

charges men (so-called “topiles”) with the task of protecting the security of its

people, although none of these are armed. The assembly, a counsel of elders,

or an indigenous tribunal hear the conflicts and impose sanctions.

Those who refuse to perform their cargos or services are normally sanc-

7Traditionally women were excluded from many of these duties. Field research revealsthat since the introduction of conditional cash transfer programs in 1997 (Progresa/nowOportunidades), women got more deeply engaged in the life of the community, and nowserve in health and education committees sponsored by the program. Women are alsoincreasingly performing other important “cargos”, but evidence from our qualitative fieldwork also suggests that it is usually not high-level responsibilities.

15

tioned. Traditionally, sanctions included imprisonment, disruption in the sup-

ply of water, physical punishment8, or in extreme cases, expulsion from the

community. In our ethnographic work we found evidence of a gradual aban-

donment of the latter sanctions. We also found that servicio is intimately

linked with cross-border dynamics. For example, some migrants will return

to their home-towns specifically to perform servicio for free, and consume a

substantial amount of their savings, reinforcing a cycle of decapitalization.

To sum up, all municipal governments in Mexico share the same formal

institutions per constitutional mandate. After the 1995 constitutional state

amendment in Oaxaca, municipalities were given the choice to switch to a

traditional system of local governance. This system is loosely organized around

the concept of usos y costumbres, but it encompasses a panoply of practices

and rules to select leaders and solve collective-choice dilemmas. Once collective

decisions are taken according to these informal rules, they are expressed in the

formal apparatus of the municipal government.

Usos y costumbres is thus a system of informal rules and practices which

are embedded in the formal, constitutional rules of municipal operation, and

provides a rich example of the multiple nature of institutions (Greif, 2006).

This embeddedness provides a unique research design opportunity because

formal institutions remain fixed while the informal practice of day to day

governance vary. Formal rules are fixed within the municipality (regarding

for example the nature and powers of municipal authorities) and they are the

same across municipalities in how they relate fiscally or legally to the state

or the federation. Despite these similarities, there is variation within Oaxaca

in the way formal authority is selected and exercised: parties and electoral

8Our interviews attributed the gradual abandonment of physical punishment to the pres-ence of the Human Rights Commissions, which received complains of human rights violationsfrom members of the communities.

16

competition in some municipalities, and assembly-based, participatory rules

in others. We exploit this variation to generate inferences about the e!ects

of governance types on participation, provision of public goods and women

participation in government.

Our theoretical expectation is that traditional governance institutions will

be able to better provide local public goods because of their superior capacity

to hold municipal presidents accountable, engage citizens in collective decision

making, and monitor and sanction non-cooperative behavior. These theoretical

expectations build from the work by Olson (1971) on the structural charac-

teristics of groups that can foster the individual provision of public goods (i.e.

monitoring and sanctioning) and the more recent extension by Ostrom (2009)

on governance of natural resources, which shows the conditions under which

communities can sometimes device informal direct participatory practices and

monitoring and adjudication devices to successfully administer public goods.

The complex social organization of Usos seems to have gradually evolved be-

cause of its functional role in allowing indigenous communities to better coor-

dinate collective action toward the provision of public goods, often in the face

of marginalization, poverty, and adversity.

4 Propensity Score Matching using Aggregate Data

Our analysis assesses the advantages and disadvantages of traditional gover-

nance relative to the institutions of representative democracy prevailing today.

Given that traditional authorities are more likely to be present in extremely

poor, indigenous and isolated communities, we make use of a matching pro-

cedure to create an appropriate counterfactual through which equally poor,

isolated, indigenous communities can be compared in their forms of gover-

17

nance.

The fundamental problem that needs to be addressed is that usos is not

randomly assigned. In fact, when the constitutional reform was carried out

in 1995, indigenous communities in Oaxaca were asked to select whether they

wanted to be ruled by one form of governance or the other. Recondo (2007)

suggests that this selection was relatively autonomous, and although there

were some adjustments in the numbers, as some municipalities shifted between

governance regimes, since 2000 the communities ruled by usos have remained

fixed at 418. Given the lack of records of the decision-making process that led

communities to choose one or the other governance system, we make use of

geography and long-term settlements patterns in order to find a counterfactual

of what the social outcome would be of making collective decisions under

systems of governance based on political parties, rather than usos. We rely on

Propensity Score matching.

The greatest challenge with the validity of propensity score is to make sure

that the balancing property is satisfied. Although this is not strictly a test

of whether non-confoundedness actually holds, when the treated and control

groups within the common support are balanced we can be more confident

that the groups are actually comparable. In the end, however, we must recog-

nize that propensity score methods are matching on observable independent

variables, so we are subject to what Ho, Imai, King, and Stuart (2007) call the

propensity score tautology, namely that the method depends on believing that

we have successfully estimated a probability of treatment that keeps the same

ordering as the unobserved true propensity score. Exact and nearest neigh-

bor matching have some advantages over propensity scores, particularly when

dealing with data that falls within few categories in its heterogeneity. But as

soon as observations can be characterized by several meaningful variables that

18

distinguish their di!erences, and the di!erences are continuous rather than

discrete, there is a curse of dimensionality typical of nonparametric estimation

methods, which render the implementation o these other methods not viable.

Figure 1 provides a sense of the spatial distribution of municipalities in

Oaxaca governed through usos and political parties. The figure presents mu-

nicipalities governed by usos in color; while the ones governed by political

parties are lightly shaded. The map also shows the municipal boundaries and

the size of localities within the municipalities, where the circle diameter cor-

responds to the relative size. Two features become apparent from the map:

usos is less prevalent in coastal areas and close to highways and large cities;

and there is a spatial clustering (suggesting spatial dependence) in the distri-

bution of usos across the territory. Although the figure does not depict this

additional feature, municipalities ruled by usos are more likely to be observed

in mountainous areas.

Figure 1: Spatial distribution of usos

The propensity score we calculate makes use of these geographic features,

19

the spatial dependence among observations, and the geographic distance of

usos communities to features such as highways and large cities. Specifically,

we calculate the propensity score as the predicted value of a probit estimation

including the following variables:

Lag Usos. Usos municipalities are geographically clustered. This is a spatial lag

indicating whether neighboring municipalities are ruled by usos. It is constructed

through a weighting matrix that captures (queen) contiguity of order two, which

means that a municipality is compared with the surrounding municipalities as well

as those that surround the (first order) neighbors in the immediate vicinity 9.

Territorial Concentration. Within each municipality, we calculated a Hirschman-

Herfindahl index of concentration, which takes into account how scattered or con-

centrated the population is in localities within the political jurisdiction. Population

at the locality level comes from the 2000 census, as reported by National Institute

for Statistics and Geography (INEGI in Spanish), using localities with at least 5

inhabitants. This variable ranges from 1 (where all the population in the munici-

pality is concentrated in only one locality) to a theoretical value of 0 (denoting total

dispersion).

Distance City. The Euclidean distance (straight line) in kilometers between the

centroid of each municipality and the closest city with more than 100,000 inhabitants

according to data from the 2000 census published by INEGI 10.

Distance Road. The Euclidean distance in kilometers between the centroid of each

municipality and the closest paved road. Roads and their quality come from INEGI.

Indigenous. The presence of indigenous traditions is a precondition for the sub-

sistence of traditional authorities. Unfortunately Mexico does not have a count of

indigenous peoples based on self-adscription. We use the percentage of the munici-

pality inhabitants over 5 years old who speak an indigenous language, as reported by

9The spatial lag was calculated using the statistical software GeoDA developed by Anselin(1998).

10Distances for this and the next variable were calculated using the ArcGIS software fromESRI.

20

the 2000 census by INEGI. This measure is consistent with recent conceptual clar-

ifications in the literature on ethnic identity. Chandra (2006) argues that a proper

conceptualization of ethnic identity requires to focus on descent-based features (ge-

netic or cultural). Ethnic markers are the subset of those descent-based identifiers

which are visible and costly to change in the short run.

Indigenous languages are reproduced inter-generationally within the household.

Not all indigenous people speak an indigenous language, but it is reasonable to

assume that people who speak an indigenous language are highly likely to be indige-

nous, specially given the fact that indigenous languages are not taught in Mexico’s

formal education system –except for special schools targeted to communities that

are already indigenous–. Therefore, indigenous languages in Mexico fall within the

constrained descent-based category proposed by Chandra, and our measure would

be a lower bound of more complete measures of ethnicity11.

Income. This variable is meant to capture structural characteristics of moderniza-

tion. Given that the correlation between municipal per capita GDP and the share

of indigenous population is extremely high (p=.47), we use the residuals of a regres-

sion of municipal GDP on indigenous. Per capita GDP is obtained from the UNDP

Human Development Report for Mexico, and is calculated on the basis of data from

the 2000 census.

Religious Fractionalization. Communities can be divided among ethnic, linguis-

tic or religious lines. Trejo (2009) has shown that the most relevant dimension of

division among communities in Southern Mexico is related to the inroads of non-

catholic Christian faiths. We measure a Hirschman-Herfindahl index of fractional-

ization with the share of religious faiths according to the 2000 census.

Altitude. The average altitude of a municipality measured in kilometers 12.

11Provided independence across categories.12Calculated using ArcGIS on the basis of raster files from INEGI.

21

5 Results

Table 1 provides the propensity score 13 using these variables. It is important to

note that the propensity score does not include dependent variables of interest,

such as coverage in the provision of public goods, political participation, or

social change.

Table 1: Propensity Score (Dependent Variable: usos)

Lag Usos 4.50***(.22 )

Territorial Concentration 1.09***(0.26 )

Distance City .0039**(0.0017 )

Distance Road 0091(0.0071 )

Indigenous 1.02***(.19 )

Income (Residuals) -8.44***(1.37 )

Religious Frag. .81*(.45 )

Altitude .37***(.11 )

Constant -4.48(.37 )

n 2382!

2 1737.98Pseudo-R2 .79

Almost all the variables are statistically significant, and the fit in the es-

timation is very good. The specific signs are reasonable, given what we know

about the historical processes that led to the retention of traditional methods

of governance in indigenous communities. The positive sign of the territorial

concentration variable suggests that usos is more prevalent in communities

13Calculated using the pscore package for Stata (Becker and Ichino, 2002).

22

concentrated in one core town. Remote municipalities farther from cities are

more likely to retain traditional authorities. Municipalities that have kept

their linguistic distinctiveness according to the indigenous variable are more

likely to keep usos. Richer places as denoted by the Income variable are likely

to be governed by political parties. Finally, higher places in the mountains are

more likely to be ruled by traditional methods.

Table 2 presents a test of balancing in the propensity score over the com-

mon support. The table reports tests of means tests between the treatment

and the control group (usos vs. parties) in each of 7 optimal blocks where

the propensity score is not statistically di!erent between groups, for each of

the correlates. The table reports the di!erence in means and the standard

deviation. None of the di!erences is statistically significant at the 99 percent

level.

Table 2: Propensity Score (Dependent Variable: usos)

Blocks 1 2 3 4 5 6 7

Lag Usos -0.1095 -0.0614 0.0398 0.0242 0.0649 -0.0142 -0.0763(0.0435 ) (0.0627 ) (0.0369 ) (0.0356 ) (0.0352 ) (0.0353 ) (0.0433 )

Fragmentation -0.0426 0.1617 0.0355 -0.1567 0.0701 0.0056 -0.0119(0.0948 ) (0.1202 ) (0.0674 ) (0.0957 ) (0.0937 ) (0.1169 ) (0.1721 )

Distance City -12.98 3.052 6.242 26.861 -9.888 -2.81 7.462(16.62 ) (17.987 ) (11.542 ) (13.204 ) (10.757 ) (12.09 ) (11.699 )

Distance Road 5.09 2.921 -0.338 -1.144 -2.337 -1.466 -11.326(4.04 ) (2.215 ) (2.085 ) (2.055 ) (2.236 ) (4.338 ) (6.012 )

Indigenous 0.1242 -0.1066 0.0066 0.0647 -0.1201 0.0961 -0.1064(0.1211 ) (0.1289 ) (0.0876 ) (0.0928 ) (0.1224 ) (0.1459 ) (0.2035 )

Income (residual) -0.0168 -0.0025 0.0173 0.002 0.0255 0.0077 -0.0537(0.0175 ) (0.0236 ) (0.0143 ) (0.0159 ) (0.0181 ) (0.0201 ) (0.0327 )

Religious Frag. 0.0334 -0.0838 -0.0285 0.0451 0.0273 -0.0121 0.0056(0.0589 ) (0.0431 ) (0.0396 ) (0.0407 ) (0.0469 ) (0.0517 ) (0.0878 )

Altitude 0.2399 0.4574 -0.2523 -0.2958 -0.1657 -0.0196 -0.2876(0.2512 ) (0.2584 ) (0.1563 ) (0.2246 ) (0.1821 ) (0.1927 ) (0.2277 )

Treated 12 16 32 16 11 8 4Control 140 13 36 46 65 65 176Observations 152 29 68 62 76 73 180

The propensity score provides useful counterfactuals to compare matched

places ruled by usos with others ruled by political parties. The 1995 politi-

cal reform allowed municipalities to switch away from a party-based system

23

in which the PRI used to control virtually all municipalities in the state of

Oaxaca.

When the reform was carried out, some municipalities were perhaps already

governed by traditional customs de facto. In those cases the reform might not

have fully changed the system of government, but provided legal certitude and

backing to the traditional practices. More importantly for our own purposes,

the reform allowed indigenous communities to retain autonomy and insulate

decision-making in their municipalities from party factionalism by eliminat-

ing party elections altogether. Thus, our relevant comparison is one between

autonomous indigenous communities versus non-autonomous ones.

Thus, we can think of usos as a treatment that may have had some e!ects

regarding social structure, political participation, the provision of public goods

and other municipal characteristics. When available we will make matching

tests using variables in first di!erences, comparing the variable of interest

before and after the reform of usos was adopted (or traditional practices were

validated by the state constitution).

Table 3 reports kernel density matching average treatment e!ects on the

treated (ATT) on a series of socio-economic and public good provision variables

14 using bootstrapped standard errors. The propensity score is recalculated in

each matching exercise in order to ensure that the uncertainty surrounding

the estimation of the propensity score is also incorporated into the calculation

of the bootstrapped errors. The variables of interest are all first di!erences

between the census indicators of 1990 and 2000. Hence the table measures

whether the treatment of usos as a governance structure had a statistically

significant e!ect on the evolution of social characteristics and the provision of

public goods. The variables are measures of religious fractionalization; changes

14As calculated with the attk routine in Stata

24

in percentages of the population who are bilingual; who speaks an indigenous

language; and who are illiterate; poverty as defined as percent of households

earning less than one minimum wage; and coverage of public goods –electricity,

water and sewerage –as measured by percentages of households with access to

these services.

Table 3: Socio-economic and Public Good Provision Matching

Change 2000-1990 ATT SE tReligious Fractionalization -0.002 0.008 -0.283Bilingual -0.007 0.013 -0.535Indigenous 0.001 0.006 0.039Illiteracy 0.008 0.011 0.719Extreme Poverty -0.08 0.036 -2.245Electricity 0.125 0.036 3.457Water -0.056 0.139 -0.4Sewerage -0.155 0.274 -0.565Treated = 412Control = 228

In terms of social development, there is not evidence suggesting lower lit-

eracy rate changes in municipalities ruled by usos. Social change as denoted

by bilingualism or speaking an indigenous language does not seem to be af-

fected by usos, as compared to its counterfactual of relatively isolated, poor

indigenous communities. Furthermore, there is no evidence of greater divi-

sion in places ruled by political parties, as could be denoted by an increase in

religious cleavages.

The ATTs are are small and statistically not significant, except for the

case of electricity provision and the extreme poverty headcount. Hence, there

is evidence suggesting that the change in the poverty rate, as measured by

the percentage of households living with earnings of less than one minimum

wage, has favored places ruled by usos. Furthermore, usos places seem to

25

be more e!ective in finding ways to extend the coverage of electricity more

widely among their inhabitants. We attribute this finding to higher equality

in the distribution of public goods in indigenous communities ruled by usos.

The provision of electricity requires first that the locality is connected to the

federal grid, but the determination of who gets electricity within the town is

determined by local governance choices.

Table 4: Socio-economic and Public Good Provision Matching

ATT SE tDi!erence margin -0.011 0.035 -0.321Di! Number of Parties 0.142 0.093 1.533Di!erence PRI -0.008 0.024 -0.33Di!erence PAN 0.011 0.015 0.692Di!erence PRD -0.013 0.02 -0.671

Former Mayor -0.011 0.022 -0.497Sessions -1.423 4.513 -0.315Open Sessions 2.136 1.137 1.879Female Council Members -0.052 0.019 -2.727Female mayors -0.025 0.012 -2.119Treated = 412Control = 228

Table 4 makes a similar exercise for variables related to political partic-

ipation. It reports the di!erence between the federal presidential electoral

patterns in 1994 and 2000 at the municipal level15. The first five variables

are measured in first di!erences, testing whether electoral competition at the

federal level was depressed, or parties became entrenched due to the usos re-

form. They report respectively the ATT for the di!erences in the margin of

victory, the e!ective number of parties, the share of the PRI, the PAN and

the PRD. None of these are statistically significant. If anything, the signs

suggest a slight increase in competitiveness as denoted by the e!ective number

15It should be noted that the choice of usos did not limit partisan political competitionfor posts at higher levels of government.

26

of parties. Hence, our results strongly suggest that there is no evidence of an

entrenchment of PRI politicians, as has been generally assumed by scholars

such as Benton (2008) or Recondo (2007). usos municipalities exhibit very

similar levels of party competition at the federal level than party municipal-

ities, which means that the PAN, the PRD, and other political parties have

been able to make similar inroads in these impoverished communities than in

the ones that allow political parties to contest for municipal elections. The

results thus support the view that the PRI is not more entrenched in these

communities

The last five rows show results regarding civic participation in collective

decision-making and whether there might be entrenchment or capture by local

elites. We use data from the Survey to Municipal Presidents on Social Devel-

opment 16. The first item asks whether the current mayor has previously held

the post. Although there is no immediate reelection in Mexico, it is not un-

common for entrenched caciques to be mayors again after some terms. There

is no evidence that in usos the same mayor has held o"ce more frequently,

and in fact, from the anthropological literature what emerges is more a sense

of rotation in charges, including the duty of mayor.

The next variables are all related to political engagement at the municipal

level. The Social Development survey asks mayors how often they carry out

council meetings, and whether those are sessions open to citizens. Although

there is no di!erence in the average number of council meetings, in munici-

palities ruled by usos those meetings are open to citizens more often. Data

from our own survey research in the state of Oaxaca reported below further

corroborates that in usos municipalities there is ample civic engagement in col-

lective decision-making whereas in party municipalities citizens remain frankly

16Encuesta Nacional para Presidentes Municipales Sobre Desarrollo Social in Spanish,carried out by the Secretariat of Social Development in 2002

27

disengaged.

Finally, the table reports two indicators of governance where usos does not

fare as well as places ruled by political parties. These are related to female

participation in the top echelons of municipal government. The variables mea-

sure the share of the municipal council made up by women, and whether the

current mayor is a woman, as reported by the National Municipal Information

System 17 in 2002. The ATT suggest that in usos there is a smaller participa-

tion of women in high o"ce. This is in keeping with the findings of Eisenstadt

(2007).

6 Individual Level Analysis of Civic Engagement and

the Distribution of Public Services

The aggregate data gives us a rough approximation to di!erences in gover-

nance and public good provision between both types of municipalities. We

can confidently reject the hypothesis that traditional governance serves to en-

trench autocratic enclaves associated with the former ruling party. We also

can confidently conclude that traditional governance does not lead to under-

provision of public goods in indigenous communities; if anything, the results

with aggregate data suggest the opposite.

However, individual-level data is needed to explore citizen engagement in

collective decision-making and its impact on the distribution of public services

within the community. We designed and collected a stratified random sample

involving 600 questionnaires of men and women over 18 years old in rural and

semi-rural areas in Oaxaca.

The stratification divided Oaxaca municipalities according to the size of

17Sistema Nacional de Informacion Muncipal (SNIM) produced by the Secretariat of Gov-ernment.

28

their cabecera (the core town, seat of the municipal government) and their

governance institutions. The conventional classification from Mexico’s statis-

tical agency, INEGI, defines a municipality as either rural or urban, according

to a threshold of whether the majority lives in localities below or above 2500

inhabitants. Such classification does not take into account the peculiar settle-

ment patterns characterizing indigenous municipalities. Indian communities

in Mexico are the result of the reconstruction of the surviving indigenous pop-

ulations after the decimation of their numbers during the 16th century. The

Spanish Crown coerced the scattered Indian inhabitants into living in urban

areas, creating pueblos de indios.

Almost every cabecera in the around 2450 Mexican municipalities today

was a pueblo during the colonial period, but there were more than 4,500 pueb-

los of the colonial era, so only half of them achieved the status of heading a

municipality of their own. The typical settlement pattern of an indigenous

municipality is to have a political jurisdiction over a relatively small land area,

with most of the population living in the cabecera. This characteristic dictated

the type of stratification we designed.

Strata were established classifying municipalities as rural when there were

less than 2500 inhabitants in the cabecera, regardless of the overall size of the

municipality; urban municipalities were those with cabeceras of between 2500

and 5000 inhabitants; and we excluded large metropolitan areas. This coverage

was established in order to enable inferences about governance conditions in

territories that could meaningfully be thought of as single units. This excluded

mid-size cities as well as some municipalities adjoining metropolitan areas.

On the governance dimension, strata were established depending on whether

municipal governments are elected trough political parties or through the use

of customary law (usos). The classification used comes from the Instituto

29

Electoral Estatal de Oaxaca, which established in 2009 that 418 municipalities

were governed through usos. The joint stratification by size and governance

yields an oversampling of rural localities and municipalities governed by usos.

These are places where much of the rural poverty is concentrated and municipal

administrations are very fragile in terms of sta"ng, skills and finance.

For the selection of individual municipalities we used the mapping of the

Federal Electoral Institute that divides the territories into tracts comprising

750 voters. Each of this so-called secciones electorales are equally likely in

terms of their probability of being chosen, hence providing a very reliable

sample frame for the kind of rural areas where the survey was carried out.

Around half of the sample points fell into cabeceras while the rest in smaller

localities called agencias. 16 questionnaires were collected in the cabeceras

and 8 in the other localities. These numbers were chosen by calculating the

power of the tests across governance structures and localities that we planned

to make, using the Optimal Design software (Spybrook et al., 2008). 60 was

too expensive due to the dispersion of the sample; 40 not enough power. The

final distribution of surveys in Usos/Parties and Cabeceras/Agencias in the

sample is depicted in figure 2.

The large dispersion of the sample is a consequence of taking a rural ap-

proach in the design of the sample. The survey was collected in August 2009.

Two survey teams had 4 routes collected. Permission was obtained from munic-

ipal presidents to enter the communities. In many cases of Usos municipalities

this meant the calling of an assembly, after which the research team was able

to visit the town or village. Below we present some basic cross-tabs from our

survey that convey the profound di!erences in civic engagement in both types

of municipalities.

Table 5.- Basic Descriptives for Usos and Parties.

30

Parties Usos

Have you ever gone to acommunity assembly of themunicipality (for usos)/ anassembly of the Cabildo (forparties)?

Yes 10.37 62.25

(25) (254)

No 89.63 37.75

(216) (154)

Does the municipal presidentconsult people to hear theirproblems and decide which publicworks he will realize?

No 62.36 31.61

(111) (122)

Yes 37.64 68.39

In general, do you think that thecurrent municipal presidentgoverns to service the communityor he does not care aboutpeople?s problems?

To service theCommunity 48.86 68.78

(86) (249)

He does notcare about ourproblems 51.14 31.22

(90) (113)

Would you say that the currentmunicipal president has carriedout public works that benefitpeople like you or not?

Yes 54.4 62.73

(99) (234)

No 45.6 37.27

(83) (139)

How much do you get to knowabout what is decided in themunicipality?

Nothing 81.48 60.4

(154) (241)

A lot 18.52 39.6

(35) (158)

The data suggests that municipalities governed by usos have significantly

higher civic engagement – citizens tend to participate in the collective decision

making, feel that their leaders are more accountable, and have more infor-

mation about government decisions. The data also suggest profound citizen

disengagement in municipalities governed by political parties citizens do not

participate in collective decision-making and have very little information about

pubic decisions. Although these di!erences seem to be quite large, they might

be driven by problems of selection and omitted variables. Below we address

31

these issues in a more systematic way.

We are interested in assessing patterns of civic engagement in usos and

party municipalities –who participates in collective decision-making and who

remains disengaged. To model civic engagement, we implement a Heckman

correction model to address the problem of selection into either type of gov-

ernance structure (Heckman, 1979). A Heckman correction allows us to take

into account the potential selection bias that may alter the estimates, and by

doing a separate regression for each governance type using the appropriate

inverse Mills ratios, to allow variation in the regression coe"cients by gover-

nance type. This enables us to explore the e!ects of governance type on each

of these coe"cients.

The dependent variable is an index of participation and civic engagement.

We used household level-data to build an index that captures the level of civic

engagement among the population. For Usos, the index captures whether

people find out about municipal government activity through town assemblies,

whether they actually attend these assemblies, and how many times they do

so in a year. The equivalent questions for municipalities governed by parties

register whether people find out about municipal government decisions through

cabildo meetings, if they attend, and how many times a year. In both cases,

the index is the simple average of each one of the standardized items that part

of it. This index is used as our primary dependent variable at this stage of the

analysis.

The first step in the models is a probit that will explain selection into gov-

ernance type, and the second step regression is a GLS with clustered standard

errors in which the clusters are the 48 municipalities where the survey was

carried out. Estimates of the covariates are made for each government type.

The selection equation is not calculated only for the 48 municipalities in our

32

−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5

0.0

0.5

1.0

1.5

Civic Engagement Index

Index

Density

PartiesUsos

survey sample frame, but for the 570 municipalities in Oaxaca. We use the

same variables as in the propensity score matching parametric estimation from

the previous sections. But given that we only include the state of Oaxaca, we

do not use a spatial lag. For the second stage we use individual level vari-

ables from the survey, and include the inverse mills ratio calculated from the

first stage, in order to control for selection e!ects. These estimations were

implemented in R.

In keeping with the literature on political engagement, participation de-

pends on individual socio-demographic characteristics, political attitudes, par-

ticipation in national social programs, as well as environmental conditions at

the community level. The community level variables are primarily driven by

the di!erences in governance structure between municipalities ruled by usos

and those selecting authorities through party competition. We expect to have

selection e!ects in participation, as well as di!erent coe"cients for those vari-

ables according to the governance structure. In particular, we hypothesize that

33

participation is influenced by the following factors:

Age. Younger members of the community will participate less than older

adults who are heads of household and have children. Other things equal, we

expect participation to taper o! after some age, where the oldest community

members are less likely to participate than when they were in the prime of

their adult lives. We operationalize this with the variable age, measured in

years, and its square (age2 ).

Gender. Given that Oaxaca is a relatively traditional society, we expect

women to participate less than men, but the e!ect will be mediated by age,

given a generational shift as younger women have become more empowered.

We therefore include the dummy variable women and its interaction with age

measured in years: women*age.

Education. We expect more engagement by the more educated citizens. This

is operationalized with three dummy variables, noeducation for those who have

no formal schooling, secondary for those who have completed some years of

middle school, highschool for those with some years of high school and college

for those with some university schooling. The baseline is for respondents with

some elementary school.

Poverty. We expect poorer households to be less engaged. We created the

dummy variable poor, indicating respondents in households falling below a

$120 USD a month poverty line.

Ethnic Identity. We operationalize indian identity with the direct question

of whether the respondent speaks an indian language (indian).

Migration. We expect migrants to be more engaged in community issues.

Oaxaca has a long tradition of international migration to the United States.

Migrants are entrepreneurial individuals who continue sending remittances and

return because they have a particular stake in their communities. We opera-

34

tionalize this as a dummy variable (migrant).

Remittances. Households that receive remittances from abroad might be less

likely to participate because they are less dependent from the community for

the provision of services, which they might opt to provide privately. We use a

dummy variable remittances.

Pay Taxes. We use the response as to whether the respondent pays property

tax as a proxy that indicates both homeownership and the contribution an

individual makes to the municipal co!ers. This is a dummy variable denoted

by the name of the property tax in Mexico, predial. We expect respondents

who pay predial to participate more.

Participation in Social Programs. We expect recipients of the Conditional

Cash Transfer program Oportunidades to be more engaged in community mat-

ters. The variable Oportunidades denotes whether the respondent family is a

recipient of the program, regardless of the fact that the cash transfer itself is

handed out only to women.

Partisan identification. Although usos entails not allowing political parties

to nominate candidates for the mayor’s o"ce, partisan politics still exists in all

Oaxacan municipalities for federal and state elections. Given the dominance

of the PRI, we expect that there will be more participation among citizens

who identify with that party. Non-identifiers to any political party might be

less inclined to participate, because they are less connected with the political

organizations. We expect any of these e!ects to be much larger in places ruled

by political parties compared to usos. The variables in the estimation are

dummies for PAN, PRD and no identification (panid, prdid and noid), where

the base category is being identified with the Parties.

Table 6.- Logit for Participation Index.

Usos Parties

35

Female 0.2427 4.1857 ***

0.8953 1.4605

Female*Age "0.0459 *** "0.1384 ***

0.0176 0.0390

No School "0.5174 "7.2984 ***

0.5784 1.1537

Middle 0.8364 * "1.0077

0.4558 1.0057

High School 0.5065 0.9740

0.7234 0.6754

College 0.3091 2.3603 **

0.7392 1.0527

Age 0.1458 *** 0.3136 ***

0.0465 0.1058

Age2 "0.0010 ** "0.0021 ***

0.0005 0.0008

Poor 0.2984 0.6708

0.3410 0.8682

No Party ID 0.3108 "3.3220 **

0.4281 1.3911

PAN 1.0677 ** "4.3638 ***

0.5521 0.9368

PRD "0.0390 "2.2520 **

0.5274 0.9721

Other 1.6478 ** "10.1465 ***

0.6871 1.5586

CCT 0.8685 ** 0.7238

0.3895 0.7712

Migrate 0.5996 * 1.4194 *

0.3297 0.8564

Property Tax "0.4914 "0.5953

0.3702 1.0782

Remiittances "0.4757 "0.5386

0.3260 1.1294

Indigenous 0.2046 "0.0565

0.3258 0.8698

IMR "1.2629 *** 1.2297

0.3022 0.9002

36

We highlight the main findings. First, even after controlling for selection

e!ects, there is considerably more civic engagement in usos municipalities as

indicated by the negative and strongly significant sign of the IMR. There is

significantly more participation of women in party municipalities, although

considering how little citizens participate in party municipalities, we do not

find evidence that women are significantly more disempowered in usos munic-

ipalities relative to party ones. Third, the poor and uneducated are seriously

disengaged in party municipalities; not such income gap exists in usos munic-

ipalities, where citizens of all income levels participate in collective decision-

making. As expected, middle age individuals participate more than the young

and old and this is true for both governance types.

With respect to partisanship, we find that propensity to participate in party

municipalities is strongly correlated with partisanship. In these municipalities,

only PRI voters tend to participate in collective decision-making, whereas those

who support other parties remain profoundly disengaged. Not such result

emerges in usos municipalities. In fact, contrary to the prevailing approach

about usos being an autocratic enclave of the PRI, those who identify with

the PAN and other parties participate more in these municipalities.

Lastly, with respect to social programs and migration, we find that Oportu-

nidades has an important role in mobilizing citizens to participate in collective

decision-making mainly in usos municipalities. A final result is that returning

migrants are more prone to participate in the community’s collective decision

making processes, and this is true both for usos and party municipalities.

The last task of the paper is to estimate the e!ects of local governance

on the distribution of public goods within the community. Presumably, the

intra-community distribution of local public goods should be in part shaped

by civic engagement. Our expectation is that, given that collective decision-

37

making in party municipalities seems to be captured by a small party elite,

the distribution of public services in these communities should happen along

along clientelistic lines. By contrast, given participatory democracy practices

in usos municipalities permit a broader group of citizens to engage in collective-

decision making, we expect to find a more egalitarian intra-community distri-

bution of public goods services.

Our survey asked respondents to tell us where they access water –whether

their household had running water in versus collected water in rivers, lakes,

or public wells and if they had sewerage. We estimate Heckman correction

models for access to these services, controlling for similar sociodemographics

and various measures of wealth.

Table 7.- Distribution of Public Goods by Governance Type (includes In-verse Mills Ratio).

Water Sewage

Usos Parties Usos Parties

Constant "1.3795 ** "0.7518 "3.7296 *** "9.3332 ***

0.7509 1.3794 1.0721 3.1539

No school "1.3203 0.6859 "8.1128 *** "1.8139

0.8951 0.8285 0.9073 2.2828

Middle 0.2399 0.9766 ** 0.7186 2.2843 ***

0.3664 0.4763 0.5416 0.8383

High School"0.2407 0.6438 "0.4065 2.6428 ***

0.6167 0.7357 0.6330 0.9402

University 0.0293 0.5781 "0.4436 1.8190 **

0.9039 0.5937 0.9611 0.8611

Poor 0.0971 "0.7407 ** 0.0599 "0.7855

0.3302 0.3396 0.3528 0.5036

No party ID"0.1354 "0.7493 ** 0.0181 "1.5496 **

0.3186 0.3523 0.3958 0.6356

PAN "0.1179 "0.2745 "0.1422 "1.2220

0.4253 0.5201 0.5932 1.1450

PRD 0.0529 "0.9066 ** 0.4550 "1.6525 **

0.4406 0.4156 0.5898 0.7127

Other "0.5417 0.9354 * 1.0500 4.2054

38

0.7343 0.5668 1.8119 0.7103

CCT "0.1763 "0.3346 "0.8517 * 2.5022 *

0.4666 0.4881 0.4657 1.4689

Migrate 0.7932 * "1.2621 "0.6912 * 0.6634

0.4874 1.2231 0.4176 0.6547

Property tax"0.0188 "0.5243 1.1727 ** 0.8683

0.4183 0.5541 0.5544 1.0868

Remittances"0.2172 0.4493 "0.7708 ** "5.7409 ***

0.3238 0.9319 0.3268 1.2328

Indigenous 1.1134 *** 1.0474 * 1.5184 *** 0.9644

0.3961 0.5481 0.5928 0.7355

Dirt floor 0.0116 "1.2515 0.5576 "2.2334 **

0.5614 0.7975 0.4818 1.0272

cabecera "0.7833 1.5639 ** 1.8657 ** 6.4989 ***

0.4849 0.7140 0.7744 2.5914

IMR 2.1476 *** "1.3424 "2.0005 *** 3.1080 *

0.6796 0.8302 0.8066 1.7467

Light bulbs 0.0827 0.0460 0.1130 ** 0.1873 **

0.0582 0.0585 0.0556 0.0788

Our results support our contention about diverging patterns of intra-community

public service distribution. The distribution of public services is more egali-

tarian or universal in usos municipalities relative to party ones. In particular,

we find that party municipalities favor in the distribution of both water and

sewage higher educated individuals (middle school and up) over those with

elementary or no schooling. By contrast, usos municipalities seem to favor

lower educated individuals or those with elementary schooling. Similarly, we

find that the poor are more often excluded from public services in party mu-

nicipalities and not such e!ect exists in usos municipalities.

Our results further indicate, consistent with out theoretical expectations,

that party municipalities politicize the distribution of services, predominantly

favoring individuals who belong to the party clientele. That is, those who

belong to the PRI (the base category) have better access to public services,

39

much better certainly than PRD but also PAN voters. None of the party

coe"cients were statistically significant in the usos municipalities, suggesting

that not such clientelistic practices could be detected in these localities.

The e!ect of indigenous language is positive for usos municipalities both in

the water and sewage regressions, suggesting that usos municipalities reward

those who retain indigenous traditions. Interestingly, we found a positive e!ect

for indigenous also in party municipalities, but only in the water regression.

Lastly, we find a strong bias in favor of cabeceras in party localities, wherein

access to both water and sewage tend to be disproportionately concentrated

in the municipality’s political center. For usos municipalities, we also find a

pro-cabecera bias in the distribution of sewage, not water, although it should

be noted that this bias is significantly stronger in party municipalities.

Overall, our findings suggest that traditional governance and direct partic-

ipatory democracy practices in indigenous communities leads to higher civic

engagement in collective decision making and more egalitarian distributions of

public services. By contrast, in similarly poor communities, citizens remain

profoundly apathetic and disengaged from collective decision making where

they are governed by political parties. Governance through political parties

allow for elite capture wherein public services tend to be distributed according

to clientelistic lines and to neglect the poor.

7 Discussion

Previous research focusing on the political dimension of the provision of public

goods recognizes three main reasons for variation in the provision of public

goods: ethnic diversity (or more in general, social heterogeneity), state ca-

pacity and local government accountability. Our analysis contributes to this

40

research agenda by focusing on the e!ects of local governance structures (di-

rect participatory democracy embedded in customary law vs representative

democracy) on the provision of public goods. The evidence we have presented

points to a less pessimistic view about the e!ects of local traditional governance

systems in indigenous Mexico, contrary to the claim of other observers.

Our hypothesis was that municipalities governed by usos would have higher

rates of change in the provision of public goods because of the political pro-

cesses that lead to the allocation of municipal funds to collective projects.

Systems of governance based on electoral competition among political parties

di!er essentially from usos because decisions are taken by politicians with-

out an on-going process of consultation to the citizenry. The monitoring and

sanctioning dynamics that come into play when citizens gather in public as-

semblies are usually absent in party-run municipalities, and thus the allocation

of resources for public goods seems sub-optimal.

These findings are corroborated in a household level survey we conducted in

the state of Oaxaca in 2010, in which we find higher levels of civic engagement

and political participation and a more egalitarian distribution of public goods

within the communities governed by usos.

The di!erences between the two types of governance that we presented in

our typology point to a broader distinction between types of democracy. The

format of decision making in electoral democracies seems to bear a higher risk

of agency loss than participatory democracies.

Although modern societies are, by their size and complexity, not the ideal

locus of direct and participatory democracy, there are lessons to be extracted

from the fact that, with regards to the provision of some basic well-being

services, non-partisan political arrangements seem to present some advantages

over the widespread party-based democratic activity. Our findings confirm

41

old ideas about the importance of participation and collective monitoring of

authority to maximize collective well-being.

A related line of thought is the profound crisis of political parties across

countries, specially in new democracies (Mainwaring et al., 2006; Hagopian

and Mainwaring, 2005; Smith, 2005; Levitsky, 2005). When new democracies

cannot provide a minimum basis of material well-being, support for democracy

erodes substantially, opening the door to anti-systemic leaders and in extreme

cases to authoritarian setbacks. The quality of new democracies seems directly

linked to the perception of government capacity to provide minimal well-being.

Public goods are an integral part of this function.

Rethinking the way electoral and party-based systems of governance work,

and contrasting their logic with other forms of participatory democracy (of

which usos is a good example) should also prove valuable in the ongoing e!ort

of institutional design to which all democracies are subject.

A policy prescription that emerges from these results leads to reevaluate the

possibility to establish usos beyond Oaxaca. We are in the process of extending

our field survey with a second out-of-sample one in which the sample frame

involves counterfactual polling points, matched to the polling points already

visited in the first survey, using propensity scores to pick locations outside of

the state of Oaxaca, where municipalities are not allowed to use traditional

governance methods.

Our goal will be to explore the impact of the formalization of usos. In

sampling indigenous communities beyond Oaxaca, we will be able to iden-

tify places that informally use traditional collective decision-making processes,

but formally are ruled by political parties, and places that are only ruled by

political parties. Hence, we will be able to compare three forms of munici-

pal governance –pure party governance, pure traditional governance, and dual

42

governance– and to analyze how these impact collective decision-making, civic

engagement, and public good provision in poor communities in Mexico.

43

References

Acemoglu, D., S. Johnson, and J. Robinson (2001). The Colonial Originsof Comparative Development: An Empirical Investigation. American Eco-nomic Review 91 (5), 1369–1401.

Alesina, A., R. Baqir, and W. Easterly (1999). Public Goods and EthnicDivisions. Quarterly Journal of Economics 114 (4), 1243–1284.

Alesina, A. and E. La Ferrara (2000). Participation in Heterogeneous Com-munities. Quarterly Journal of Economics 115 (3), 847–904.

Anselin, L. (1998). OpenGeoDa.

Baqir, R. (2002). Social Sector Spending in a Panel of Countries. Workingpaper 02/35, International Monetary Fund.

Bardhan, P. and D. Mookherjee (2000). Corruption and Decentralization ofIinfrastructure Delivery in Developing Countries. Boston University, Insti-tute for Economic Development.

Becker, S. and A. Ichino (2002). Estimation of Average Treatment E!ectsBased on Propensity Scores. The Stata Journal 2 (4), 358–377.

Benton, A. (2008). Bottom-up Challenges to National Democracy: Mexicos(legal) Subnational Authoritarian Enclaves. Comparative Politics .

Bernhard, H., U. Fischbacher, and E. Fehr (2006). Parochial Altruism inHumans. Nature 442 (7105), 912–915.

Besley, T. (2006). Principled agents?: The Political Economy of Good Gov-ernment. Oxford University Press.

Besley, T. and R. Burgess (2002). The Political Economy of GovernmentResponsiveness: Theory and Evidence from India. Quarterly Journal ofEconomics 117 (4), 1415–1451.

Besley, T. and S. Coate (2003). Centralized versus Decentralized Provisionof Local Public Goods: A Political Economy Approach. Journal of publiceconomics 87 (12), 2611–2637.

Besley, T. and A. Prat (2001). Handcu!s on the Grabbing Hand? Media Cap-ture and Government Accountability. Mimeo, London School of Economics .

Chandra, K. (2006). What is Ethnic Identity and Does It Matter? AnnualReview of Political Science 9.

Chattopadhyay, R. and E. Duflo (2004). Women as Policy Makers: Evidencefrom a Randomized Policy Experiment in India. Econometrica 72 (5), 1409–1443.

44

Chhibber, P. and I. Nooruddin (2004). Do Party Systems Count? ComparativePolitical Studies 37 (2), 152.

Cleary, M. (2004). Electoral Competition and Democracy in Mexico. Ph. D.thesis.

Cleary, M. (2007). Electoral Competition, Participation, and GovernmentResponsiveness in Mexico. American Journal of Political Science 51 (2),283–299.

Dayton-Johnson, J. (2000). Determinants of Collective Action on the LocalCommons: A Model with Evidence from Mexico. Journal of DevelopmentEconomics 62 (1), 181–208.

Diaz-Cayeros, A., B. Magaloni, and F. Estevez (2007). Private versus Pub-lic Goods as Electoral Investments: A Portfolio Diversification Model withEvidence from Mexico. In H. Kitschelt and S. Wilkinson (Eds.), Patrons,Clients, and Policies. Patterns of Democratic Accountability and PoliticalCompetition. Cambridge University Press.

Eisenstadt, T. (2007). Usos y Costumbres and Postelectoral Conflicts in Oax-aca, Mexico, 1995-2004: An Empirical and Normative Assessment. LatinAmerican Research Review 42 (1), 52–77.

Fearon, J. and D. Laitin (2003). Ethnicity, Insurgency, and Civil War. Amer-ican Political Science Review 97 (01), 75–90.

Ferraz, C. and F. Finan (2005). Reelection Incentives and Political Corruption:Evidence from Brazilian Audit Reports. University of California at Berkeley(Working paper).

Glaeser, E., D. Laibson, J. Scheinkman, and C. Soutter (2000). MeasuringTrust. Quarterly Journal of Economics 115 (3), 811–846.

Greif, A. (2006). Institutions and the Path to the Modern Economy: Lessonsfrom Medieval Trade. Cambridge University Press.

Habyarimana, J., M. Humphreys, D. Posner, and J. Weinstein (2007). WhyDoes Ethnic Diversity Undermine Public Goods Provision? American Po-litical Science Review 101 (04), 709–725.

Hagopian, F. and S. Mainwaring (2005). The Third Wave of Democratizationin Latin America: Advances and Setbacks. Cambridge University Press.

Ho, D. E., K. Imai, G. King, and E. A. Stuart (2007). Matching as nonpara-metric preprocessing for reducing model dependence in parametric causalinference. Political Analysis 15 (3), 199–236.

45

Khwaja, A. (2009). Can Good Projects Succeed in Bad Communities? Journalof Public Economics 93 (7-8), 899–916.

Kohli, A. (2001). The Success of India’s Democracy. Cambridge UniversityPress.

Levitsky, S. (2005). Argentina: Democratic Survival amidst Economic Failure.In F. Hagopian and S. P. Mainwaring (Eds.), The Third Wave of Democ-ratization in Latin America: Advances and Setbacks. Cambridge UniversityPress.

Mainwaring, S., A. Bejarano, and E. Leongomez (2006). The Crisis of Demo-cratic Representation in the Andes. Stanford University Press.

Miguel, E. (2004). Tribe or Nation?: Nation Building and Public Goods inKenya versus Tanzania. World Politics 56 (3), 327–362.

Miguel, E. and M. Gugerty (2005). Ethnic Diversity, Social Sanctions, andPublic Goods in Kenya. Journal of Public Economics 89 (11-12), 2325–2368.

Olken, B. (2006). Corruption and the Costs of Redistribution: Micro Evidencefrom Indonesia. Journal of Public Economics 90 (4-5), 853–870.

Olson, M. (1971). The Logic of Collective Action: Public Goods and the Theoryof Groups. Harvard University Press.

Ostrom, E. (2009). A General Framework for Analyzing Sustainability ofSocial-Ecological Systems. Science 325 (5939).

Platteau, J. (2004). Monitoring Elite Capture in Community-Driven Develop-ment. Development and Change 35 (2), 223–246.

Recondo, D. (2007). La Polıtica del Gatopardo : Multiculturalismo y Democ-racia en Oaxaca. Mexico D.F.: CIESAS. Publicaciones de la Casa Chata.

Reinikka, R. and J. Svensson (2004). Local Capture: Evidence From a Cen-tral Government Transfer Program in Uganda. Quarterly Journal of Eco-nomics 119 (2), 679–705.

Sen, A. (1999). Development as Freedom. Oxford University Press.

Smith, P. (2005). Democracy in Latin America: Political Change in Compar-ative Perspective. Oxford University Press.

Trejo, G. (2009). Religious Competition and Ethnic Mobilization in LatinAmerica: Why the Catholic Church Promotes Indigenous Movements inMexico. American Political Science Review 103 (03), 323–342.

World Bank (2003). World Development Report 2004: Making Services Workfor Poor People. Technical report.

46